To address some of the problems in existing distracted driving behavior detection methods, such as low detection accuracy and poor real-time performance, a deep learning-based target detection method was used for driver distracted driving behavior detection. Firstly, a distracted driving behavior dataset was constructed, including images of drivers using mobile phones, drinking water and smoking, and the targets were annotated, secondly a lightweight target detection model NanoDet was selected for training and validation. The results show that the method can accurately and quickly identify driver behaviors including using mobile phones, drinking water and smoking while driving.

This paper proposed a path planning method with low computing power requirement and small parking space for path planning of vertical parking space. First, the one-step and multi-step base path planning for vertical parking were designed based on the linear-arc combination considering the requirements of collision constraints and vehicle kinematic constraints in the parking process. Then, the curvature optimization was combined with the clothoid curve to realize the curvature smoothing of the parking path. Finally, the feasibility of the method was verified by simulation. The results show that the method can plan a safe and smooth parking path for different lateral parking spaces and different initial attitude angles.

To provide the typical hazardous distracted driving scenarios for the development and testing of vehicle active safety systems, this research relied upon an in-depth investigation of 375 distracted driving accidents, determined parameters of the scenarios across three dimensions including road environment, participant’s speed, and motion state, compared and analyzed national statistics with samples from each accident type and extracted key feature parameters. Using the two-step cluster analysis method, this research obtained typical distracted driving scenarios of 11 different accident types associated with distracted driving, these scenarios were further refined to derive 4 core test scenarios by integrating key feature parameters.

In order to improve the low lane departure warning rate due to the poor robustness of the traditional edge operator in lane feature extraction and the weak fitting ability of the traditional Hough transform curve, this paper proposed a lane departure warning method based on the threshold segmentation of the optimized maximum inter-class variance method (OTSU algorithm) and the sliding window method. Firstly, the genetic annealing algorithm was used to optimize and solve the optimal threshold of OTSU algorithm, and the Holistically-nested Edge Detection (HED) model was invoked to obtain the edge features of lane lines, and the area of interest was converted into an aerial view. Then, the sliding window method was utilized to slice the lane lines and the second-order polynomial fitting was carried out for the lane pixels in the window one by one. Finally, the lane departure warning and curve warning were given according to the relative position of the vehicle and the lane line. The test results show that the accuracy of the proposed method is 95.92%, and the detection rate can reach 34 ms/frame.

For the situation that the current automatic parking system requires a high number of sensors and computing power, this paper proposed an automatic parking system based on panoramic images and human-machine interaction. Quantitative perception training was performed on the improved Vacant Parking Slot Network (VPS-Net) to realize real-time parking slot detection and parking slot occupancy classification. At the same time, with the help of the driver’s judgment on the surrounding environment, only 4 surround-view fisheye cameras around the car body were used to complete unoccupied parking slot detection and real-time monitoring of the parking environment, and multi-stage path planning and path-following controller were utilized to realize smooth and accurate parking of the vehicle into the parking slot. The verification results show that the system can realize automatic parking under various typical parking slots and lighting conditions.

This paper elaborated the overall framework of commercial vehicle fatigue early warning system, and analyzed in details the research status of monitoring system, human machine interface and fatigue detection method. The paper pointed out that the future monitoring system should have the ability of high stability, short delay and massive data processing, and divided the warning into two levels, and defined the effective human-computer interaction respectively. The paper then analyzed 4 categories of fatigue detection methods, and indicated that the fatigue detection method based on multi-feature information fusion will be the main research direction in the future. The paper finally revealed the difficulties of the current research, and prospected research of the commercial vehicle fatigue warning system from 3 aspects, i.e. obtaining more driver information, extracting more fatigue features, and reducing the dependence on specific fatigue features.

A curve braking force control strategy was developed for vehicles equipped with Electro Mechanical Brake (EMB) systems. Firstly, the required braking force was obtained according to the driver’s expected deceleration, and then the braking force was initially distributed based on the vertical load estimate, and then the additional yaw moment control module designed based on the Active Disturbance Rejection Control (ADRC) algorithm was utilized to obtain the additional yaw moment required to improve handling stability of the vehicle, and finally the initially assigned braking force was adjusted through the braking force adjustment module. Simulink and CarSim were applied to co-simulate and compare with the proportional distribution scheme, and the results show that the four wheels are not easy to lock in curve braking when the control strategy proposed in this research is used, and the yaw rate and sideslip angle are closer to their ideal values, which effectively improves the safety of cornering braking.

For the influence of electromagnetic coupling vibration caused by random road excitation and unbalanced electromagnetic excitation of in-wheel motor on electric vehicle driven by in-wheel motor, this paper proposed a vibration suppression method based on Active Disturbance Rejection Control(ADRC). Firstly, the mathematical model of a quarter of the vehicle vibration system was established. Secondly, a controller based on ADRC was designed. The road random excitation and the electromagnetic excitation generated by the motor itself were regarded as the total disturbance for unified observation compensation control. Finally, the proposed control strategy was verified by MATLAB/Simulink. The results show that ADRC can suppress the electromagnetic coupling vibration caused by road roughness and eccentric electromagnetic force.

To address the problems of small dataset and low efficiency of artificial diagnosis method in the research process of passenger vehicle abnormal noise recognition, this paper proposed an efficient intelligent recognition technique, which applied data expansion method with high recognition accuracy and adopted the parallel working mechanism of Convolutional Neural Network (CNN) and Transformer encoder stack to obtain the classification model. It is found that the data expansion method can effectively improve the classification performance when the extracted Mel Frequency Cepstral Coefficients (MFCCs) features of the augmented data are used as the input to the parallel network, and the proposed model can achieve classification accuracy up to 98.31% on the testing dataset.

In order to meet the urgent demand of water-gas separation at the anode hydrogen outlet of the vehicle fuel cell stack and in view of the problems of large space occupation and low water-gas separation efficiency, this paper proposed a new box-type dual-baffle water gas separation device design based on the principle of baffle separation. Through CFD and finite element modeling and simulation methods, combined with the actual working conditions of hydrogen recirculation in a 60 kW fuel cell system, the separation efficiency of the device was simulated and analyzed. The influence of key factors such as inlet and outlet diameter, front and rear baffle length, distance from the inner wall, inclination angle, arc radius and angle on the separation efficiency was studied, and the optimal structural parameters of the device was determined which obviously improved the separation efficiency.